论文部分内容阅读
考虑到构建二叉树支持向量机时样本的分布情况对分类器推广能力具有较大影响,提出一种改进的二叉树支持向量机层次结构构建方法.以类间样本距离和带权值的类内样本距离与其标准差的比值作为类的分类度.将类间距离大且类内样本平均分布广的类最先分离.利用标准数据集,通过与不同多类分类算法比较,验证了改进的二叉树支持向量机的优越性.对双转子涡喷发动机气路部件进行应用改进的算法进行故障诊断,得到了较好的故障识别率.
Considering that the distribution of samples when constructing binary tree SVM has a great influence on the generalization ability of the classifier, an improved hierarchical construction method of binary tree SVM is proposed.It takes the inter-class sample distance and weighted intra-class sample distance And the standard deviation of the ratio as the class of classification.Classification of the distance between the class is large and the average class of the samples are widely separated class separation.Using standard data sets, compared with the different classification algorithms to verify the improved binary tree support vector The superiority of the machine.An improved algorithm for the gas components of the twin-rotor turbojet engine is used to diagnose the fault, and a better fault recognition rate is obtained.